2007/10/9, Martijn van Oosterhout <[EMAIL PROTECTED]>:
> On Tue, Oct 09, 2007 at 06:40:23PM +0200, Pavel Stehule wrote:
> > It needs always seq scan :(, and take space on buffer cache. Solution
> > based on random generated PK are much faster. I collaborate with one
> > my customer. He shows random products from 10K products on every page
> > of one eShop. And he cannot understand, so ORDER random() LIMIT is bad
> > trick, because this trick is on PostgreSQL FAQ.
>
> It's the only trick that works in all situations though. There are
> ofcourse faster methods, but they require information about the
> distribution of the values, the type, PKs, indexes etc...
>
> The standard does have stuff relating to extracting samples from
> tables, but they're not implemented.
>
I agree. I don't wont to remove it from FAQ. I would to add note, so
sometimes is necessary to find other trick.
Regards
Pavel
---------------------------(end of broadcast)---------------------------
TIP 9: In versions below 8.0, the planner will ignore your desire to
choose an index scan if your joining column's datatypes do not
match
